Stable Diffusion overview in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

How does Stable Diffusion generate images?

Stable Diffusion creates images by starting with random noise and gradually refining it. What is the main process it uses to turn noise into a clear image?

A It copies images from a database based on keywords.

B It directly draws images pixel by pixel without any noise.

C It uses a diffusion process that removes noise step-by-step guided by a learned model.

D It uses a simple random guess and checks if the image looks good.

💡 Hint

Think about how noise is changed gradually to form an image.

Stable Diffusion starts with random noise and uses a trained neural network to remove noise gradually, producing a clear image step-by-step.

❓ Model Choice

intermediate

Which model type is Stable Diffusion based on?

Stable Diffusion uses a specific kind of neural network architecture to generate images. Which one is it?

A Transformer-based U-Net

B Convolutional Neural Network (CNN)

C Recurrent Neural Network (RNN)

D Generative Adversarial Network (GAN)

💡 Hint

It combines attention mechanisms with convolution layers.

Stable Diffusion uses a U-Net architecture enhanced with transformer attention layers to better capture image details during denoising.

❓ Metrics

advanced

What metric is commonly used to evaluate image quality in Stable Diffusion outputs?

When checking how good the images generated by Stable Diffusion are, which metric is often used to measure similarity to real images?

A Mean Squared Error (MSE)

B Cross-Entropy Loss

C Accuracy

D Fréchet Inception Distance (FID)

💡 Hint

This metric compares distributions of generated and real images in a feature space.

FID measures how close the distribution of generated images is to real images, capturing quality and diversity.

❓ Hyperparameter

advanced

Which hyperparameter controls the trade-off between creativity and faithfulness in Stable Diffusion?

Stable Diffusion uses a parameter to balance how closely the output matches the prompt versus how creative or varied the image is. What is this parameter called?

A Guidance scale

B Learning rate

C Batch size

D Number of diffusion steps

💡 Hint

It adjusts how strongly the model follows the text prompt.

The guidance scale controls how much the model sticks to the prompt versus exploring creative variations.

🔧 Debug

expert

What error occurs if the noise schedule is incorrectly set in Stable Diffusion?

If the noise schedule (the way noise is added and removed) is set incorrectly in Stable Diffusion, what kind of problem will most likely happen during image generation?

A The model will run out of memory and crash immediately.

B The model will produce images with random noise and no clear structure.

C The model will generate images instantly without any noise steps.

D The model will produce perfectly clear images every time.

💡 Hint

Think about what happens if noise is not removed properly.

An incorrect noise schedule means the model cannot properly denoise, resulting in noisy, unclear images.